Automatic model-based roentgen stereophotogrammetric analysis (RSA) of total knee prostheses

Conventional radiography is insensitive for early and accurate estimation of the mal-alignment and wear of knee prostheses. The two-staged (rough and fine) registration of the model-based RSA technique has recently been developed to in vivo estimate the prosthetic pose (i.e, location and orientation). In the literature, rough registration often uses template match or manual adjustment of the roentgen images. Additionally, possible error induced by the nonorthogonality of taking two roentgen images neither examined nor calibrated prior to fine registration. This study developed two RSA methods for automate the estimation of the prosthetic pose and decrease the nonorthogonality-induced error. The predicted results were validated by both simulative and experimental tests and compared with reported findings in the literature. The outcome revealed that the feature-recognized method automates pose estimation and significantly increases the execution efficiency up to about 50 times in comparison with the literature counterparts. Although the nonorthogonal images resulted in undesirable errors, the outline-optimized method can effectively compensate for the induced errors prior to fine registration. The superiority in automation, efficiency, and accuracy demonstrated the clinical practicability of the two proposed methods especially for the numerous fluoroscopic images of dynamic motion.     

Assessment of the effect of mesh density on the material property discretisation within QCT based FE models: a practical example using the implanted proximal tibia

A three-dimensional, quantitative computed tomography based finite element model of a proximal implanted tibia was analysed in order to assess the effect of mesh density on material property discretisation and the resulting influence on the predicted stress distribution. The mesh was refined on the contact surfaces (matched meshes) with element sizes of 3, 2, 1.4, 1 and 0.8 mm. The same loading conditions were used in all models (bi-condylar load: 60% medial, 40% lateral). Significant variations were observed in the modulus distributions between the coarsest and finest mesh densities. Poor discretisation of the material properties also resulted in poor correlations of the stresses and risk ratios between the coarsest and finest meshes. Little difference in Young's modulus, von Mises stress and risk ratio distributions were observed between the three finest models; hence, it was concluded that for this particular case an element size of 1.4 mm on the contact surfaces was enough to properly describe the stiffness, stress and risk ratio distributions within the bone. Poor convergence of the material property distribution occurred when the element size was significantly larger than the pixel size of the source CT data. It was concluded that unless there is convergence in the Young's modulus distribution, convergence of the stress field or of other parameters of interest will not occur either.     

Subject specific finite element mesh generation of the pelvis from biplanar x-ray images: application to 120 clinical cases

Several Finite Element (FE) models of the pelvis have been developed to comprehensively assess the onset of pathologies and for clinical and industrial applications. However, because of the difficulties associated with the creation of subject-specific FE mesh from CT scan and MR images, most of the existing models rely on the data of one given individual. Moreover, although several fast and robust methods have been developed for automatically generating tetrahedral meshes of arbitrary geometries, hexahedral meshes are still preferred today because of their distinct advantages but their generation remains an open challenge. Recently, approaches have been proposed for fast 3D reconstruction of bones based on X-ray imaging. In this study, we adapted such an approach for the fast and automatic generation of all-hexahedral subject-specific FE models of the pelvis based on the elastic registration of a generic mesh to the subject-specific target in conjunction with element regularity and quality correction. The technique was successfully tested on a database of 120 3D reconstructions of pelvises from biplanar X-ray images. For each patient, a full hexahedral subject-specific FE mesh was generated with an accurate surface representation.     

A new model-based RSA method validated using CAD models and models from reversed engineering

Roentgen stereophotogrammetric analysis (RSA) was developed to measure micromotion of an orthopaedic implant with respect to its surrounding bone. A disadvantage of conventional RSA is that it requires the implant to be marked with tantalum beads. This disadvantage can potentially be resolved with model-based RSA, whereby a 3D model of the implant is used for matching with the actual images and the assessment of position and rotation of the implant. In this study, a model-based RSA algorithm is presented and validated in phantom experiments. To investigate the influence of the accuracy of the implant models that were used for model-based RSA, we studied both computer aided design (CAD) models as well as models obtained by means of reversed engineering (RE) of the actual implant. The results demonstrate that the RE models provide more accurate results than the CAD models. If these RE models are derived from the very same implant, it is possible to achieve a maximum standard deviation of the error in the migration calculation of 0.06 mm for translations in x- and y-direction and 0.14 mm for the out of plane z-direction, respectively. For rotations about the y-axis, the standard deviation was about 0.1 degrees and for rotations about the x- and z-axis 0.05 degrees. Studies with clinical RSA-radiographs must prove that these results can also be reached in a clinical setting, making model-based RSA a possible alternative for marker-based RSA.     

Assessment of factors influencing finite element vertebral model predictions

This study aimed to establish model construction and configuration procedures for future vertebral finite element analysis by studying convergence, sensitivity, and accuracy behaviors of semiautomatically generated models and comparing the results with manually generated models. During a previous study, six porcine vertebral bodies were imaged using a microcomputed tomography scanner and tested in axial compression to establish their stiffness and failure strength. Finite element models were built using a manual meshing method. In this study, the experimental agreement of those models was compared with that of semiautomatically generated models of the same six vertebrae. Both manually and semiautomatically generated models were assigned gray-scale-based, element-specific material properties. The convergence of the semiautomatically generated models was analyzed for the complete models along with material property and architecture control cases. A sensitivity study was also undertaken to test the reaction of the models to changes in material property values, architecture, and boundary conditions. In control cases, the element-specific material properties reduce the convergence of the models in comparison to homogeneous models. However, the full vertebral models showed strong convergence characteristics. The sensitivity study revealed a significant reaction to changes in architecture, boundary conditions, and load position, while the sensitivity to changes in material property values was proportional. The semiautomatically generated models produced stiffness and strength predictions of similar accuracy to the manually generated models with much shorter image segmentation and meshing times. Semiautomatic methods can provide a more rapid alternative to manual mesh generation techniques and produce vertebral models of similar accuracy. The representation of the boundary conditions, load position, and surrounding environment is crucial to the accurate prediction of the vertebral response. At present, an element size of 2x2x2 mm(3) appears sufficient since the error at this size is dominated by factors, such as the load position, which will not be improved by increasing the mesh resolution. Higher resolution meshes may be appropriate in the future as models are made more sophisticated and computational processing time is reduced.     

Comparative study on the performance of textural image features for active contour segmentation

We present a computerized method for the semi-automatic detection of contours in ultrasound images.The novelty of our study is the introduction of a fast and efficient image function relating to parametric active contour models.This new function is a combination of the gray-level information and first-order statistical features,called standard deviation parameters.In a comprehensive study,the developed algorithm and the efficiency of segmentation were first tested for synthetic images.Tests were also performed on breast and liver ultrasound images.The proposed method was compared with the watershed approach to show its efficiency.The performance of the segmentation was estimated using the area error rate.Using the standard deviation textural feature and a 5×5 kernel,our curve evolution was able to produce results close to the minimal area error rate(namely 8.88% for breast images and 10.82% for liver images).The image resolution was evaluated using the contrast-to-gradient method.The experiments showed promising segmentation results.     

The effects of different mesh generation methods on computational fluid dynamic analysis and power loss assessment in total cavopulmonary connection

The flow field and energetic efficiency of total cavopulmonary connection (TCPC) models have been studied by both in vitro experiment and computational fluid dynamics (CFD). All the previous CFD studies have employed the structured mesh generation method to create the TCPC simulation model. In this study, a realistic TCPC model with complete anatomical features was numerically simulated using both structured and unstructured mesh generation methods. The flow fields and energy losses were compared in these two meshes. Two different energy loss calculation methods, the control volume and viscous dissipation methods, were investigated. The energy losses were also compared to the in vitro experimental results. The results demonstrated that: (1) the flow fields in the structured model were qualitatively similar to the unstructured model; (2) more vortices were present in the structured model than in the unstructured model; (3) both models had the least energy loss when flow was equally distributed to the left and right pulmonary arteries, while high losses occurred for extreme pulmonary arterial flow splits; (4) the energy loss results calculated using the same method were significantly different for different meshes; and (5) the energy loss results calculated using different methods were significantly different for the same mesh.     

The use of sparse CT datasets for auto-generating accurate FE models of the femur and pelvis

The finite element (FE) method when coupled with computed tomography (CT) is a powerful tool in orthopaedic biomechanics. However, substantial data is required for patient-specific modelling. Here we present a new method for generating a FE model with a minimum amount of patient data. Our method uses high order cubic Hermite basis functions for mesh generation and least-square fits the mesh to the dataset. We have tested our method on seven patient data sets obtained from CT assisted osteodensitometry of the proximal femur. Using only 12 CT slices we generated smooth and accurate meshes of the proximal femur with a geometric root mean square (RMS) error of less than 1 mm and peak errors less than 8 mm. To model the complex geometry of the pelvis we developed a hybrid method which supplements sparse patient data with data from the visible human data set. We tested this method on three patient data sets, generating FE meshes of the pelvis using only 10 CT slices with an overall RMS error less than 3 mm. Although we have peak errors about 12 mm in these meshes, they occur relatively far from the region of interest (the acetabulum) and will have minimal effects on the performance of the model. Considering that linear meshes usually require about 70-100 pelvic CT slices (in axial mode) to generate FE models, our method has brought a significant data reduction to the automatic mesh generation step. The method, that is fully automated except for a semi-automatic bone/tissue boundary extraction part, will bring the benefits of FE methods to the clinical environment with much reduced radiation risks and data requirement.     

Requirements for mesh resolution in 3D computational hemodynamics

Computational techniques are widely used for studying large artery hemodynamics. Current trends favor analyzing flow in more anatomically realistic arteries. A significant obstacle to such analyses is generation of computational meshes that accurately resolve both the complex geometry and the physiologically relevant flow features. Here we examine, for a single arterial geometry, how velocity and wall shear stress patterns depend on mesh characteristics. A well-validated Navier-Stokes solver was used to simulate flow in an anatomically realistic human right coronary artery (RCA) using unstructured high-order tetrahedral finite element meshes. Velocities, wall shear stresses (WSS), and wall shear stress gradients were computed on a conventional "high-resolution" mesh series (60,000 to 160,000 velocity nodes) generated with a commercial meshing package. Similar calculations were then performed in a series of meshes generated through an adaptive mesh refinement (AMR) methodology. Mesh-independent velocity fields were not very difficult to obtain for both the conventional and adaptive mesh series. However, wall shear stress fields, and, in particular, wall shear stress gradient fields, were much more difficult to accurately resolve. The conventional (nonadaptive) mesh series did not show a consistent trend towards mesh-independence of WSS results. For the adaptive series, it required approximately 190,000 velocity nodes to reach an r.m.s. error in normalized WSS of less than 10 percent. Achieving mesh-independence in computed WSS fields requires a surprisingly large number of nodes, and is best approached through a systematic solution-adaptive mesh refinement technique. Calculations of WSS, and particularly WSS gradients, show appreciable errors even on meshes that appear to produce mesh-independent velocity fields.     

The mesh-matching algorithm: an automatic 3D mesh generator for finite element structures

Several authors have employed finite element analysis for stress and strain analysis in orthopaedic biomechanics. Unfortunately, the definition of three-dimensional models is time consuming (mainly because of the manual 3D meshing process) and consequently the number of analyses to be performed is limited. The authors have investigated a new patient-specific method allowing automatically 3D mesh generation for structures as complex as bone for example. This method, called the mesh-matching (M-M) algorithm, generated automatically customized 3D meshes of anatomical structures from an already existing model. The M-M algorithm has been used to generate FE models of 10 proximal human femora from an initial one which had been experimentally validated. The automatically generated meshes seemed to demonstrate satisfying results.     

Smooth surface meshing for automated finite element model generation from 3D image data

Finite element (FE) modelling based on data from three-dimensional high-resolution computed tomography (CT) imaging systems provides a non-invasive method to assess structural mechanics. Automated mesh generation from these voxel based image data can be achieved by direct conversion to hexahedron elements, however these model representations have jagged edges. This paper proposes an automated method to generate smoothed FE meshes from voxel-based image data. Mesh fairing processes are utilized that allow constraints that control the smoothing process, and are computationally efficient. Surfaces of the mesh on the exterior, as well as interfaces between two tissues, can be smoothed by varying fairing parameters and constraint criteria. The method was tested on a variety of real and simulated three-dimensional data sets, resulting in both hexahedron and tetrahedron meshes. It was shown that the fairing process is linearly related to the number of smoothing iterations, and that peak stresses are reduced in FE simulations of the smoothed models. Although developed for micro-CT data sets, this fast and reliable mesh smoothing method could be applied to any three-dimensional image data where node and element connectivity have been defined.     

Variability of measurements of cranial growth in the rabbit

This study concerns techniques used in experimental cranial growth research: roentgen cephalometry, roentgen stereophotogrammetry, and gross measurements (osteometry). A comparison of the precision of these methods has not been found in the literature. Computation of technical errors is fundamental to the sound evaluation of registered findings, and such a presentation must be obligatory in all biometric reports. We compared the measurement error of roentgen cephalometric and osteometric data with that obtained by roentgen stereophotogrammetry (RSA). RSA demonstrates a superior replicability, and this technique gives possibilities for kinematic and volumetric determinations simultaneously with distance evaluation. Roentgen cephalometry has the advantage of enabling distance and angular measurements between any well-defined skeletal points or lines. This technique, preferably after implantation of bone markers, is a reliable alternative, but optimal results necessitates calculations of the magnification factor for each bone segment involved. Direct osteometry does not contribute additional information, but problems of image magnification are omitted. Preferably, one individual should perform all measurements regardless of the method used. Growth rates and values calculated by one technique cannot be directly transformed to some other approach. In all probability, assessments of distance changes would gain substantially by using one technical approach consistently throughout actual age intervals. The least variable measurements of sutural growth are made for sutures growing primarily in one plane and with substantial growth rates. One must realize that differences among studies may be due to the limitations of, in particular, the cephalometric and osteometric techniques.     

ImageParser: a tool for finite element generation from three-dimensional medical images

Background

The finite element method (FEM) is a powerful mathematical tool to simulate and visualize the mechanical deformation of tissues and organs during medical examinations or interventions. It is yet a challenge to build up an FEM mesh directly from a volumetric image partially because the regions (or structures) of interest (ROIs) may be irregular and fuzzy.

Methods

A software package, ImageParser, is developed to generate an FEM mesh from 3-D tomographic medical images. This software uses a semi-automatic method to detect ROIs from the context of image including neighboring tissues and organs, completes segmentation of different tissues, and meshes the organ into elements.

Results

The ImageParser is shown to build up an FEM model for simulating the mechanical responses of the breast based on 3-D CT images. The breast is compressed by two plate paddles under an overall displacement as large as 20% of the initial distance between the paddles. The strain and tangential Young's modulus distributions are specified for the biomechanical analysis of breast tissues.

Conclusion

The ImageParser can successfully exact the geometry of ROIs from a complex medical image and generate the FEM mesh with customer-defined segmentation information.

     

Rapid three-dimensional segmentation of the carotid bifurcation from serial MR images

The current trend in computational hemodynamics is to employ realistic models derived from ex vivo or in vivo imaging. Such studies typically produce a series of images from which the lumen boundaries must first be individually extracted (i.e., two-dimensional segmentation), and then serially reconstructed to produce the three-dimensional lumen surface geometry. In this paper, we present a rapid three-dimensional segmentation technique that combines these two steps, based on the idea of an expanding virtual balloon. This three-dimensional technique is demonstrated in application to finite element meshing and CFD modeling of flow in the carotid bifurcation of a normal volunteer imaged with black blood MRI. Wall shear stress patterns computed using a mesh generated with the three-dimensional technique agree well with those computed using a mesh generated from conventional two-dimensional segmentation and serial reconstruction. In addition to reducing the time required to extract the lumen surface from hours to minutes, our approach is easy to learn and use and requires minimal user intervention, which can potentially increase the accuracy and precision of quantitative and longitudinal studies of hemodynamics and vascular disease.     

Automatic preparation,calibration, and simulation of deformable objects

Many simulation environments – particularly those intended for medical simulation – require solid objects to deform at interactive rates, with deformation properties that correspond to real materials. Furthermore, new objects may be created frequently (for example, each time a new patient's data is processed), prohibiting manual intervention in the model preparation process. This paper provides a pipeline for rapid preparation of deformable objects with no manual intervention, specifically focusing on mesh generation (preparing solid meshes from surface models), automated calibration of models to finite element reference analyses (including a novel approach to reducing the complexity of calibrating nonhomogeneous objects), and automated skinning of meshes for interactive simulation.     

Improving the local solution accuracy of large-scale digital image-based finite element analyses

Digital image-based finite element modeling (DIBFEM) has become a widely utilized approach for efficiently meshing complex biological structures such as trabecular bone. While DIBFEM can provide accurate predictions of apparent mechanical properties, its application to simulate local phenomena such as tissue failure or adaptation has been limited by high local solution errors at digital model boundaries. Furthermore, refinement of digital meshes does not necessarily reduce local maximum errors. The purpose of this study was to evaluate the potential to reduce local mean and maximum solution errors in digital meshes using a post-processing filtration method. The effectiveness of a three-dimensional, boundary-specific filtering algorithm was found to be mesh size dependent. Mean absolute and maximum errors were reduced for meshes with more than five elements through the diameter of a cantilever beam considered representative of a single trabecula. Furthermore, mesh refinement consistently decreased errors for filtered solutions but not necessarily for non-filtered solutions. Models with more than five elements through the beam diameter yielded absolute mean errors of less than 15% for both Von Mises stress and maximum principal strain. When applied to a high-resolution model of trabecular bone microstructure, boundary filtering produced a more continuous solution distribution and reduced the predicted maximum stress by 30%. Boundary-specific filtering provides a simple means of improving local solution accuracy while retaining the model generation and numerical storage efficiency of the DIBFEM technique.     

Strategies towards rapid generation of forefoot model incorporating realistic geometry of metatarsals encapsulated into lumped soft tissues for personalized finite element analysis

Use of finite element (FE) foot model as a clinical diagnostics tool is likely to improve the specificity of foot injury predictions in the diabetic population. Here we proposed a novel workflow for rapid construction of foot FE model incorporating realistic geometry of metatarsals encapsulated into lumped forefoot’s soft tissues. Custom algorithms were implemented to perform unsupervised segmentation and mesh generation to directly convert CT data into a usable FE model. The automatically generated model provided higher efficiency and comparable numerical accuracy when compared to the model constructed using a traditional solid-based mesh process. The entire procedure uses MATLAB as the main platform, and makes the present approach attractive for creating personalized foot models to be used in clinical studies.